The Bidirectional Gated Recurrent Unit Network Based on the Inception Module (Inception-BiGRU) Predicts the Missing...

Area of Science:

• Geophysics
• Petroleum Engineering
• Data Science

Background:

• Acoustic (AC) logging data is crucial for reservoir analysis, bridging seismic and well log information.
• Missing AC data due to equipment failure or borehole issues hinders accurate reservoir characterization.
• Re-logging is often impractical and costly, necessitating data completion methods.

Purpose of the Study:

• To develop an accurate method for completing missing AC logging data using other available logging parameters.
• To leverage deep learning for improved reservoir geological interpretation through high-resolution inversion profiles.

Main Methods:

• A hybrid deep learning model, Inception-BiGRU, was developed.
• The Inception module extracts features from logging data.
• A bidirectional gated recurrent unit network processes these features to predict missing AC data, considering sequential dependencies.

Main Results:

• The Inception-BiGRU model demonstrated higher accuracy in completing AC logging data compared to traditional models like GRU and LSTM.
• The model effectively utilizes extracted features and sequential data characteristics for accurate prediction.

Conclusions:

• The Inception-BiGRU model offers a robust and accurate solution for the challenging problem of AC logging data completion.
• This approach provides a valuable new strategy for reservoir characterization and quantitative evaluation in the presence of data gaps.